Analysis of Extreme Reversals in Seasonal and Annual Precipitation Anomalies Across the United States,

Transcription

1 Analysis of Extreme Reversals in Seasonal and Annual Precipitation Anomalies Across the United States, Michael L. Marston Thesis submitted to the faculty of Virginia Polytechnic Institute and State University in partial fulfillment of the requirements for the degree of Master of Science In Geography Andrew W. Ellis, Committee Chair Gregory B. Goodrich David F. Carroll April 18 th, 2016 Blacksburg, VA Keywords: climate variability, precipitation, Niño 3.4

2 Analysis of Extreme Reversals in Seasonal and Annual Precipitation Anomalies Across the United States, ABSTRACT Michael Marston As population and urbanization increase across the United States, the effects of natural hazards may well increase, as extreme events would increasingly affect concentrated populations and the infrastructure upon which they rely. Extreme precipitation is one natural hazard that could stress concentrated populations, and climate change research is engaging heavy precipitation frequency and its impacts. This research focuses on the less-studied phenomenon of an extreme precipitation reversal - defined as an unusually wet (dry) period that is preceded by an unusually dry (wet) period. The magnitude is expressed as the difference in the percentiles of the consecutive periods analyzed. This concept has been documented only once before in a study that analyzed extreme precipitation reversals for a region within the southwestern United States. That study found that large differences in precipitation from consecutive winters, a hydrologically critical season for the region, occurred more frequently than what would be expected from random chance, and that extreme precipitation reversals have increased significantly since This research expands upon the previous work by extending the analysis to the entire continental United States and by including multiple temporal resolutions. Climate division data were used to determine seasonal and annual precipitation for each of nine climate regions of the continental United States from Precipitation values were then ranked and given percentiles for seasonal and annual data. The season-to-season analysis was performed in two ways. The first examined consecutive seasons (e.g., winter spring, spring summer) while the second analyzed the seasonal data from consecutive years (e.g., spring 2014 spring 2015). The annual data represented precipitation for the period October 1 September 30, or the water year used by water resource managers. Following the approach of the previous study, a secondary objective of the research was to examine large-scale climate teleconnections for historical relationships with the occurrence of precipitation reversals. The El Nino-Southern Oscillation was chosen for analysis due to its well-known relationships with precipitation patterns across the United States. Results indicate regional expressions of a propensity for extreme precipitation reversals and relationships with teleconnections that may afford stakeholders guidance for proactive management. Precipitation reversal (PR) and extreme precipitation reversal (EPR) values were significantly larger for the second half of the study period for the western United States for the winter-to-winter, spring-to-spring, and year-to-year analyses. The fall-to-fall analysis also revealed changes in PR/EPR values for several regions, including the northwest, the Northern Rockies and Plains, and the Ohio Valley. Relationships between the winter-to-winter PR time series and an index representing the El Nino-Southern Oscillation (ENSO) phenomenon were examined. The winter-to-winter PR time series of the Northern Rockies and Plains region and the South exhibited significant relationships with the time series of Niño 3.4 values. El Niño (La Niña) coincided with more wet-to-dry (dry-to-wet) PR/EPR values for the Northern Rockies and Plains, while El Niño (La Niña) coincided with more dry-to-wet (wet-to-dry) PR/EPR values for the South. ii

3 Acknowledgements I would like to thank my advisor Dr. Andrew Ellis for his help and guidance throughout this process. I would also like to thank committee members Dr. Gregory Goodrich and David Carroll for their support of this research. Finally, I would like to thank the Faculty of the Department of Geography for the encouragement and help along the way. iii

4 Table of Contents List of figures. List of Tables. List of Abbreviations. vi viii xi Chapter 1: Introduction Precipitation Volatility Impacts of Precipitation Volatility Precipitation Reversals Study area 5 Chapter 2: Literature Review Hydroclimatic Research at the National Scale Regional Studies 8 a. Western United States 8 b. Central United States Conclusion 12 Chapter 3: A Climatology of Extreme Precipitation Reversals Introduction Data a Precipitation Data b Nino 3.4 Data Methods - Extreme Precipitation Reversal (EPR) Identification Results and Discussion a Year-to-Year PR and EPR Analyses b Interannual PR and EPR Analyses Analyses 28 i. Winter-to-Winter 28 ii. Spring-to-Spring 31 iii. Summer-to-Summer 34 iv. Fall-to-Fall c Intraannual PR and EPR Analyses 41 i. Winter-to-Spring 41 ii. Spring-to-Summer 44 iii. Summer-to-Fall 49 iv. Fall-to-Winter Winter-to-Winter PR/EPR relationships with Niño Conclusions 60 Chapter 4: Conclusions.. 63 iv

5 References.. 67 v

6 List of Figures Figure 1 Study Area: NCEI Climate Regions... 6 Figure 2 Study Area: NCEI Climate Regions.. 17 Figure 3 Monthly precipitation climatology for all CRs.. 19 Figure 4 Boxplots of the year-to-year PR distribution for both halves of the historical record for CR2 (West) 26 Figure 5 Year-to-Year PR/EPR time series plot for CR2 (West) Figure 6 Figure 7 Figure 8 Boxplots of the winter-to-winter PR distribution for both halves of the historical record for CR2 (West) Boxplots of the spring-to-spring PR distribution for both halves of the historical record for CR2 (West) Positive (dry-to-wet) reversals for the spring-to-spring time step plotted through time for CR7 (Southwest) Figure 9 Fall-to-Fall PR/EPR time series plot for CR5 (Ohio Valley) Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Figure 15 Figure 16 Boxplots that compare the largest 30 PR values for both halves of data of the year-to-year PR distribution for CR1 (Northwest). 39 Boxplots that compare the largest 30 PR values for both halves of data of the year-to-year PR distribution for CR3 (Northern Rockies and Plains) Boxplots that compare the largest 30 PR values for both halves of data of the year-to-year PR distribution for CR1 (Ohio Valley).. 40 Positive (dry-to-wet) reversals for the fall-to-fall time step plotted through time for CR9 (Northeast) Positive (dry-to-wet) reversals for the winter-to-spring time step plotted through time for CR1 (Northwest) Boxplots of the spring-to-summer PR distribution for both halves of the historical record for CR9 (Northeast) Boxplots that compare the largest 30 PR values for both halves of data of the spring-to-summer PR distribution for CR5 (Ohio Valley) vi

7 Figure 17 Figure 18 Figure 19 Figure 20 Figure 21 Figure 22 Figure 23 Boxplots that compare the largest 30 PR values for both halves of data of the spring-to-summer PR distribution for CR9 (Northeast) Positive (dry-to-wet) reversals for the spring-to-spring time step plotted through time for CR5 (Ohio Valley) Negative (wet-to-dry) reversals for the spring-to-summer time step plotted through time for CR7 (Southwest) 48 Negative (wet-to-dry) reversals for the spring-to-summer time step plotted through time for CR8 (Southeast). 49 Boxplots of the summer-to-fall PR distribution for both halves of the historical record for CR3 (Northern Rockies and Plains PR time series for CR8 (Southeast) of both positive (dry-to-wet) and negative (wet-to-dry) reversals for the summer-to-fall time step.. 52 Negative (wet-to-dry) reversals for the summer-to-fall time step plotted through time for CR1 (Northwest) 53 Figure 24 Fall-to-Winter PR/EPR time series plot for CR8 (Southeast) Figure 25 Figure 26 Figure 27 Figure 28 Boxplots that compare the largest 30 PR values for both halves of data of the fall-to-winter PR distribution for CR8 (Southeast) Boxplots that compare the largest 30 PR values for both halves of data of the fall-to-winter PR distribution for CR9 (Northeast) PR time series for CR8 (Southeast) of both positive (dry-to-wet) and negative (wet-to-dry) reversals for the fall-to-winter time step 57 PR time series for CR9 (Northeast) of both positive (dry-to-wet) and negative (wet-to-dry) reversals for the fall-to-winter time step 57 Figure 29 Averaged September, October, and November (SON) Niño 3.4 index time series ( ) 59 vii

8 List of Tables Table 1 Abbreviations for the nine Climate Regions.. 17 Table 2 Autocorrelation results for the year-to-year analysis Table 3 Table 4 Table 5 Table 6 Frequency per century of various EPR and PR classes for the year-to-year analysis. The Return Interval for various PR/EPR events as well as the trend of PR values for all CRs Statistical comparison of each half of the year-to-year PR time series for each of the 9 climate regions.. 26 Average of the largest 30 year-to-year PR values for each half of the historical record for the nine climate regions 28 Statistical trends of year-to-year PR values, calculated without using ABS value. 28 Table 7 Autocorrelation results for the winter-to-winter analysis Table 8 Table 9 Frequency per century of various EPR and PR classes for the winter-to-winter analysis. The Return Interval for various PR/EPR events as well as the trend of PR values for all CRs Statistical comparison of each half of the winter-to-winter PR time series for each of the 9 climate regions.. 30 Table 10 Autocorrelation results for the spring-to-spring analysis Table 11 Table 12 Table 13 Frequency per century of various EPR and PR classes for the spring-to-spring analysis. The Return Interval for various PR/EPR events as well as the trend of PR values for all CRs Statistical comparison of each half of the spring-to-spring PR time series for each of the 9 climate regions.. 33 Statistical trends of spring-to-spring PR values, calculated without using ABS value. 33 Table 14 Autocorrelation results for the summer-to-summer analysis Table 15 Frequency per century of various EPR and PR classes for the summer-to-summer analysis. The Return Interval for various PR/EPR events as well as the trend of PR values for all CRs. 35 viii

9 Table 16 Statistical comparison of each half of the summer-to-summer PR time series for each of the 9 climate regions.. 35 Table 17 Autocorrelation results for the fall-to-fall analysis Table 18 Table 19 Table 20 Table 21 Frequency per century of various EPR and PR classes for the fall-to-fall analysis. The Return Interval for various PR/EPR events as well as the trend of PR values for all CRs 36 Statistical comparison of each half of the fall-to-fall PR time series for each of the 9 climate regions.. 38 Average of the largest 30 fall-to-fall PR values for each half of the historical record for the nine climate regions 38 Statistical trends of fall-to-fall PR values, calculated without using ABS value. 40 Table 22 Autocorrelation results for the winter-to-spring analysis Table 23 Table 24 Table 25 Frequency per century of various EPR and PR classes for the winter-to-spring analysis. The Return Interval for various PR/EPR events as well as the trend of PR values for all CRs 42 Statistical comparison of each half of the winter-to-spring PR time series for each of the 9 climate regions.. 43 Statistical trends of winter-to-spring PR values, calculated without using ABS value. 43 Table 26 Autocorrelation results for the spring-to-summer analysis Table 27 Table 28 Table 29 Table 30 Frequency per century of various EPR and PR classes for the spring-to-summer analysis. The Return Interval for various PR/EPR events as well as the trend of PR values for all CRs 44 Statistical comparison of each half of the spring-to-summer PR time series for each of the 9 climate regions.. 45 Average of the largest 30 spring-to-summer PR values for each half of the historical record for the nine climate regions 46 Statistical trends of spring-to-summer PR values, calculated without using ABS value 47 Table 31 Autocorrelation results for the summer-to-fall analysis ix

10 Table 32 Table 33 Table 34 Table 35 Frequency per century of various EPR and PR classes for the summer-to-fall analysis. The Return Interval for various PR/EPR events as well as the trend of PR values for all CRs 50 Statistical comparison of each half of the summer-to-fall PR time series for each of the 9 climate regions.. 51 Average of the largest 30 summer-to-fall PR values for each half of the historical record for the nine climate regions 51 Statistical trends of summer-to-fall PR values, calculated without using ABS value 52 Table 36 Autocorrelation results for the fall-to-winter analysis Table 37 Table 38 Table 39 Table 40 Table 41 Table 42 Frequency per century of various EPR and PR classes for the fall-to-winter analysis. The Return Interval for various PR/EPR events as well as the trend of PR values for all CRs 54 Statistical comparison of each half of the fall-to-winter PR time series for each of the 9 climate regions.. 55 Average of the largest 30 summer-to-fall PR values for each half of the historical record for the nine climate regions 55 Statistical trends of summer-to-fall PR values, calculated without using ABS value 56 Pearson correlation coefficients and statistical significance (ρ) for the winter-to-winter PR/EPR calculated without absolute value and the time series and the time series of Niño 3.4 (SON) 59 The occurrences of EPR60 events for the winter-to-winter and spring-to-spring PR time series for CR2 with the phases of ENSO x

11 List of Abbreviations AMO AR5 COOP CR CR1 CR2 CR3 CR4 CR5 CR6 CR7 CR8 CR9 DJF ENSO EPR ESPR GPCP HCN IPCC NCEI NCEP NOAA PCA PR PDO SON Atlantic Multidecadal Oscillation Fifth Assessment Report Cooperative Observer Network Climate Region Northwest West Northern Rockies and Plains Upper Midwest Ohio Valley South Southwest Southeast Northeast December, January, and February El Niño-Southern Oscillation extreme precipitation reversal extreme seasonal precipitation reversal Global Precipitation Climatology Project Historical Climatology Network Intergovernmental Panel on Climate Change National Centers for Environmental Information National Center for Environmental Prediction National Oceanic and Atmospheric Administration Principal Component Analysis precipitation reversal Pacific Decadal Oscillation September, October, and November xi

12 CHAPTER 1 Introduction 1.1 Precipitation Volatility According to the National Centers for Environmental Information s (NCEI) Climatological Rankings, which is a time series of historically ranked precipitation records from ( the state of Oklahoma experienced its fifth driest October-September period on record during , followed by the 5 th wettest October-September period during The state of Virginia recorded its 2 nd driest October-September on record during , followed by the wettest October-September during Furthermore, for the October-September period during , the state of Texas had its 13 th wettest October-September period, followed by the driest October- September on record during If such reversals in precipitation are common throughout the United States, practitioners whose operations and resources are affected by seasonal-to-annual precipitation must guard against impacts from such volatility. From a broader perspective, the societal impacts may be magnified as the population distribution of the country continues to shift toward urban areas. According to the fifth assessment report (AR5) of the Intergovernmental Panel on Climate Change (IPCC), the population of the United States was 86% urbanized as of 2014, and mirroring expectations on a global scale, urbanization and population values within the United States are expected to increase in the future (IPCC 2014). The increase in population and urbanization suggests that the effects of natural hazards could increase, as singular, extreme events could affect concentrated populations and therefore a greater number of people and the infrastructure upon which they rely. Extreme precipitation is one natural hazard that could stress 1

13 dense populations of people by threatening clean water resources, testing aging infrastructure, and threatening agricultural crops that many from around the world are dependent upon. The likelihood of an increase in extreme precipitation events is put into context within the fifth assessment report (AR5) of the IPCC, which stated that almost all areas of North America exhibit very likely increases of 5 to 20% in the 20-year return value of extreme precipitation by the mid-21st-century (IPCC 2014, pg. 1456). AR5 also provided a wide-ranging outlook that states: Global warming of approximately 2 C (above the preindustrial baseline) is very likely to lead to more frequent extreme heat events and daily precipitation extremes over most areas of North America, more frequent low-snow years, and shifts toward earlier snowmelt runoff over much of the western USA and Canada. Together with climate hazards such as higher sea levels and associated storm surges, more intense droughts, and increased precipitation variability, these changes are projected to lead to increased stresses to water, agriculture, economic activities, and urban and rural settlements (high confidence) (IPCC 2014, pg. 1443). According to the AR5 report (IPCC 2013), the global temperature has risen by roughly 1 C since 1880, and conservative estimates suggest that it will likely rise another degree by Increasing temperatures are thought to increase lower tropospheric water vapor, leading to more intense wet and dry extremes depending on geographic location (Allan and Soden 2008; Held and Soden 2006; Trenberth 1999). The importance of understanding precipitation variability is further underscored by Tsonis (1996) who stated: Precipitation (rain and snow) provides, through its participation in the global hydrological and energy cycles, most of the heat flux within the atmosphere. For this reason, a knowledge of precipitation variability is important to understanding the behavior and changes of the Earth s climate system. (Tsonis 1996, pg. 700) 2

14 1.2 Impacts of Precipitation Volatility The American economy has been affected by extreme weather events through history, including floods and intense precipitation (IPCC 2014). IPCC s AR5 illuminates emerging risks, such as droughts, that could negatively affect the North American economy in the future (IPCC 2014). Two of the sectors affected by climate variability, and especially precipitation variability, are water resources and agriculture (IPCC 2014). Vörösmarty et al. (2000) noted that one fundamental concern is the impact of this climate change on water supply (pg. 284), which will likely fluctuate more than in previous decades. Extreme variability in water levels, caused in part by precipitation variability, will likely require more intensive water resource management. For example, even with extensive preparation, Phoenix, Arizona in the semi-arid southwestern United States has a looming threat of water stress if a prolonged drought were to occur (Morehouse et al. 2002). Agriculturalists may also come under increased stress as raising crops and livestock would be more difficult if precipitation variability trends toward the extreme. For example, research tying the El Niño-Southern Oscillation (ENSO) phenomenon to regional weather patterns has shown that there are economic advantages for agriculturalists who account for ENSO when making crop growing decisions (Chen et al. 2001; Hansen et al. 2001; Tack and Ubilava 2013). The effects of extreme precipitation likely are not limited to the water resources and agricultural sectors, as other economic sectors and population centers may also see negative impacts from increasing precipitation variability, including the areas of human health and ecosystem health (IPCC 2014). According to IPCC s AR5 (IPCC 2013), increased precipitation variability is expected to occur across the globe as Earth s atmosphere and surface continue to warm. The IPCC has also determined that it is likely that the frequency and intensity of heavy precipitation events over 3

15 many land areas will, on average, increase in the near-term (IPCC 2013, pg. 992). These previous statements have been provided with very limited spatial representation, as well as with limited details about whether the frequency of heavy precipitation events and the overall variability of precipitation have already begun to change. Gu et al. (2007) used the Global Precipitation Climatology Project (GPCP) dataset for a 27-year time period ( ) to resolve that while the global linear change of precipitation was near zero, there were positive and negative linear trends for regions of the globe. These linear trends are important but should not be exclusively used to determine past precipitation trends. Martino et al. (2013) examined precipitation variability using station data, and pointed out that many previous precipitation variability studies used data that ended in the mid-1990s, and analyzed small spatial extents. Martino et al. (2013) also stated that the wide variety of climate types across the United States supports its use as a robust study area for precipitation variability studies. This thesis aims to add to the baseline of knowledge regarding precipitation variability in hopes of providing greater context to climate projections by examining precipitation data across the 120-year period for the large spatial area of the contiguous United States. 1.3 Precipitation Reversals The focus of this research is extreme precipitation reversals, which are dramatic changes in precipitation anomalies (wet-to-dry or dry-to-wet) generally on relatively coarse spatial (regional) and temporal (seasonal-to-annual) scales. Several past studies of precipitation extremes have focused on individual events as opposed to seasonal and annual totals that have more of an effect on water supply. It is anticipated that a deeper understanding of extreme precipitation reversals in the face of a changing climate will allow precipitation sensitive 4

16 stakeholders to make more informed planning decisions related to shorter-term operational issues or to longer-term methodological or infrastructural changes. The work is guided by the following basic research question: Has the warming of the global climate in recent decades coincided with greater volatility in seasonal-to-annual precipitation across the United States? The answer to this question leads to applied questioning involving the degree of risk that any volatility poses for precipitation-sensitive stakeholders across the United States. While this research question is not directly engaged within this thesis, the results of this work help to better understand the value of the question going forward. This thesis is comprised of four chapters. Chapter 2 will provide a review of the research literature involving precipitation variability. Chapter 3 will provide a climatology of precipitation reversals (PR) for the nine climate regions of the contiguous United States, at multiple time steps including: year-to-year, winter-to-winter, spring-to-spring, summer-to-summer, fall-to-fall, winter-to-spring, spring-to-summer, summer-to-fall, and fall-to-winter. Lastly, Chapter 4 will summarize the work and offer concluding thoughts. 1.4 Study Area The area of study for this research is the contiguous United States, which coincides with the spatial coverage of the primary data to be used in this research. Data for this large study area will be examined at a finer resolution by using the nine climate regions across the contiguous United States as outlined by the National Centers for Environmental Information (NCEI) (Figure 1). These climate regions are comprised of finer resolution climate divisions. Climate divisions were established in the 1950s based on a combination of climatic homogeneity, drainage basin delineations, and crop types (Guttman and Quayle 1996). However, the division boundaries 5

17 ultimately follow county boundaries, and regional boundaries ultimately follow state boundaries. The United States was selected as the study area using the same rationale as that of Martino et al. (2013), who stated that the wide range of climate types across the United States makes it a good domain for precipitation studies. The United States was also selected as the study domain based on the quality and quantity of data from across the country. Figure 1: The nine climate regions of the United States (NCEI 2015) ( 6

18 CHAPTER 2 Literature Review 2.1 Hydroclimatic Research at the National Scale Several studies have aimed at improved understanding of precipitation patterns across the United States for the purpose of resolving if patterns have changed over time. Martino et al. (2013) examined precipitation variability by investigating data from 400 weather stations across the contiguous United States from The variability of precipitation was analyzed using an entropy approach. For their study, a disorder index was calculated for each station and then analyzed for temporal variability on seasonal and monthly time scales. Martino et al. (2013) also investigated spatial precipitation variability by comparing the disorder indexes of stations in close proximity to one another. It was found that temporal inconsistency in precipitation variability is higher at seasonal time scales, and that precipitation variability was uniform spatially, except for within the Mediterranean climate in California where precipitation variability displayed high spatial variability in the fall and summer seasons (Martino et al. 2013). The objectives of the study produced by Karl and Knight (1998) were to determine how precipitation had changed over time in the United States, and if changes were found, to determine how precipitation over time varied by season. This study used a variety of precipitation datasets including: the Historical Climatology Network (HCN) daily precipitation dataset, the climatological state divisional precipitation monthly dataset, and TD3200 daily precipitation dataset. Karl and Knight (1998) found that precipitation increased during the 1900s across the United States, with the most noticeable increases occurring in the spring and fall. Increases in precipitation during the winter and summer were also found, but were less 7

19 pronounced when compared to the other two seasons (Karl and Knight 1998). However, according to Karl and Knight (1998) the scale of the precipitation changes differs, up to 4% per century, depending on the dataset used. Kunkel et al. (2003) took advantage of a newly expanded precipitation dataset. This new dataset was a result of digitizing the National Weather Service s Cooperative Observer Network (COOP) pre-1948 data. The expansion of precipitation data allowed Kunkel et al. (2003) to find that there were high precipitation frequencies in the late 1800s and early 1900s, in addition to the previously known increase in precipitation frequencies that has occurred since the 1960s. When comparing the periods of increased precipitation frequencies, the late 19 th century and early 20 th century increase was found to be higher than that which has occurred since the 1960s (Kunkel et al. 2003). 2.2 Regional Studies Several studies have analyzed how precipitation is changing on a regional level. The temporal scope of these studies has varied, but a majority of the studies have taken place across the western half of the United States. Concern over water availability has been ubiquitous in the western states, while such concerns are less in the eastern states, even though the IPCC (2014) has very clearly stated that the only region in the United States where concern about water availability is unwarranted is the Pacific Northwest. 2.2a Western United States Precipitation analyses conducted by geographers and climatologists have been heavily skewed toward the west coast region of the United States. Miller and Goodrich (2007) used 8

20 climate division precipitation data to examine trends and the regionalization of winter (October- March) precipitation across the northwestern United States. They used rotated Principal Component Analysis (PCA) to create four sub-regions across the northwestern United States. ENSO and PDO data were then analyzed with the precipitation data for trends. After examining trends in precipitation for the four sub-regions, Miller and Goodrich (2007) determined that there were differences in the trends of winter precipitation for the four sub-regions, suggesting that the northwestern United States precipitation patterns may not be as spatially coherent as past research has indicated. Ellis and Barton (2012) examined a possible relationship between the north Pacific jet stream and winter (December, January, and February) precipitation variability across the western United States. The mean seasonal position and speed of the north Pacific jet stream were calculated from re-analysis data obtained from the National Center for Environmental Prediction (NCEP) and the climate division precipitation data were gathered from the National Oceanic and Atmospheric Administration (NOAA) (Ellis and Barton 2012). Ellis and Barton (2012) found that the position of the north Pacific jet stream had a significant relationship with winter precipitation for 32 out of 64 climate divisions across the western United States. As for how these particular studies relate to the proposed work, the work of Miller and Goodrich (2007) will likely provide useful information regarding the regionalization of precipitation patterns. Depending on the initial results of the proposed research, the results found by Ellis and Barton (2012) may be useful to help explain precipitation variability at a regional level, or provide methods for examining the hydroclimatic impact of atmospheric jet streams for additional regions across the United States. The only precipitation reversal analysis found in the literature was completed for the southwestern United States, and it appears that no study has examined precipitation reversals at 9

21 the national scale, as with the study proposed here. Goodrich and Ellis (2008) identified what they coined extreme seasonal precipitation reversal (ESPR) events across Arizona. The research presented in their paper sought to determine if extreme precipitation events had become more frequent in Arizona, and if so how the events might be tied to various climatic teleconnections. The methods portion of their study is important because of the similarities with the approach to the study that is presented in this thesis. Goodrich and Ellis used climate division data for the period 1895 through The data were aggregated to winter (DJF) precipitation totals and translated into percentiles. The absolute value of the percentile difference for consecutive years was used to characterize the magnitude of the year-to-year change. Going beyond the absolute value representation, Goodrich and Ellis also analyzed whether or not the change was from wet-to-dry, or from dry-to-wet. To provide context to how often different magnitudes of precipitation variability occurred historically, 1000 sets of 111 numbers (dictated by the 111 years in their study) were randomly generated. These numbers were then compared to the actual number of times extreme seasonal precipitation reversals (ESPR) occurred historically. The outcome was that the frequency of any ESPR within the historical record (independent of how extreme) was above the expected number for each of the seven climate divisions of Arizona. The number of high ESPR events (defined as occurrences of the absolute value of the difference in percentiles for consecutive years greater than 60) had also increased since Teleconnection data were also analyzed for relationships with the occurrence of ESPRs. ENSO neutral conditions coincided with roughly one-half of the ESPR events. However, Goodrich and Ellis (2008) were surprised to find that index values for the Atlantic Multidecadal Oscillation and Eastern Pacific Oscillation (EPO) were more strongly correlated to ESPRs than ENSO, regardless of the direction of the reversal (wet-to-dry or dry-to-wet). In conclusion, Goodrich 10

22 and Ellis (2008) found that the frequency of precipitation reversals had in fact increased across Arizona through the 20 th century and that the frequency with which they occurred was related historically to large-scale climate teleconnections. 2.2b Central United States One region that has not been rigorously studied is the central United States. Hu and Feng (2001) used gridded data to study interannual precipitation fluctuation across the central United States for the summer months of June, July, and August. It was found that the forcing variable for the fluctuation in precipitation actually changes every 30 to 40 years. It is important to note the temporal extent of this study was only 90 years, and this caused the authors to question their results due to the timescale of the forcing fluctuation. Another examination of precipitation variability using an entropy approach was completed for the state of Texas by Mishra et al. (2009). Their study, which was completed before the similar study by Martino et al. (2013) mentioned above, examined 43 stations across Texas from on monthly, seasonal, and annual time steps. A disorder index was calculated to measure precipitation variability over time for each station, and those indexes were then compared temporally and spatially. The results revealed that the amount of variation in annual precipitation increases from east to west across Texas. Also, Mishra et al. (2009) found that the magnitude of variation through space of precipitation variability varied across Texas for different seasons, with winter precipitation exhibiting the most spatial variability and summer the least. Woodhouse and Overpeck (1998) analyzed 2000 years of drought variability for the central United States. This study was completed using tree ring data, historical newspaper articles, and other accounts to map and temporally place the droughts of the past. Woodhouse 11

23 and Overpeck (1998) found that droughts have previously been more severe and had a much greater spatial extent than droughts in the past 150 years. Due to their findings of more serious droughts in the past, Woodhouse and Overpeck (1998) proposed that future studies consider the full range of natural interannual and interdecadal droughts, and not just the magnitudes of droughts in recent memory, when proposing possible impacts on today s population. 2.3 Conclusion An area of hydroclimatology that has been understudied is the notion of precipitation volatility and relatively rapid reversals in precipitation anomalies. With the exception of Goodrich and Ellis (2008), who analyzed ESPR events for the seven climate divisions in Arizona, no studies have examined seasonal extreme precipitation reversals. Goodrich and Ellis aimed to determine if extreme precipitation events had become more frequent in Arizona, and how the events might be tied to various climatic teleconnections. The results of their study illustrated that the frequency of ESPRs have in fact increased for much of Arizona (Goodrich and Ellis 2008). Based on their results, there is reason to hypothesize that other areas may have also experienced a change in extreme precipitation reversals. Other studies have looked at many different aspects of precipitation variability, including short-term and local variation, as well as the physical drivers of those changes. Studies have also examined how the frequencies and magnitude of precipitation amount might change under different global warming scenarios. Yet, extreme precipitation reversals surprisingly have not been studied on a broader spatial scale, and there could be great value for stakeholders in analyzing the frequency of reversals. Due to the lack of studies that have analyzed ESPR events and the importance of precipitation variability within the climate change arena as conveyed in IPCC s AR5 (2014), it appears that there is value 12

24 to performing an analysis of extreme precipitation reversals for the contiguous United Sates on varying temporal scales. While it is important to ultimately identify physical mechanisms that may help explain the cause of ESPRs and that could potentially be used to help forecast the magnitude and frequency of ESPRs, this thesis offers only a precursory analysis using one basic climate teleconnection index. 13

25 CHAPTER 3 A Climatology of Extreme Precipitation Reversals By Michael Marston Department of Geography Virginia Tech, Blacksburg, Virginia ABSTRACT Precipitation reversals are a type of natural hazard that could stress dense populations as urban areas continue to expand and grow. The research presented here provides a climatology of precipitation reversals, singularly defined as an unusually wet (dry) period that is preceded by an unusually dry (wet) period. For this research, the magnitude of precipitation reversals is defined as the difference in percentiles from one increment of data to the next. This research expands upon previous work for the winter season within the Southwestern United States by extending the analysis to the entire continental United States and by including multiple temporal resolutions. A climatology of extreme precipitation events was created for nine climate regions at multiple temporal scales using climate division precipitation data. Analyzed were reversals for consecutive seasons (e.g., winter spring, spring summer), seasons from consecutive years (e.g., summer 2014 summer 2015), and consecutive water years (October 1 September 30). The climatology illustrates regional expressions of the propensity for extreme precipitation reversals that may afford stakeholders guidance for proactive management. The west experienced significantly larger PR/EPR values for the second half of the study period for the winter-towinter, spring-to-spring, and year-to-year analyses. For the fall-to-fall analysis there were shifts in PR/EPR values over time for several regions, including a significantly decreasing trend of PR/EPR values for the Ohio Valley. The winter-to-winter time series of precipitation reversals were also compared to fall season El Nino-Southern Oscillation index values. The time series of winter-to-winter precipitation reversals for the region of the Northern Plains and Rocky Mountains and the south-central region exhibited highly correlated relationships with fall season El Nino-Southern Oscillation variability. 14

26 3.1 Introduction According to IPCC s AR5 (IPCC 2013), increased precipitation variability is expected to occur across the globe as Earth s atmosphere and surface continue to warm. As discussed in Chapter 1, precipitation volatility has the potential to impact several sectors of the United States, and several notable swings in precipitation have occurred in recent years. Several past studies have focused on precipitation volatility. Martino et al. (2013) examined precipitation variability using station data, and pointed out that many previous precipitation variability studies used data that ended in the mid-1990s and focused on small spatial extents. Martino et al. (2013) also stated that the wide variety of climate types across the United States supports its use as a robust study area for precipitation variability studies. Goodrich and Ellis (2008) examined precipitation reversals, which they defined as dramatic changes in precipitation anomalies (wet-to-dry or dry-to-wet), for the southwestern United States. The magnitude of precipitation reversals within the work of Goodrich and Ellis (2008) were determined as the difference in the percentile expression of two consecutive events. The research presented here replicates the methods of Goodrich and Ellis (2008) for defining extreme precipitation reversals, and it aims to fill the gap in the literature mentioned by Martino et al. (2013) by examining precipitation data across the 120-year period 1895 through 2014 for the large spatial area of the contiguous United States. A climatology of extreme precipitation reversal (EPR) events across the contiguous United States is presented in this chapter. It is anticipated that a deeper understanding of extreme precipitation reversals in the face of a changing climate will allow precipitation sensitive stakeholders to make more informed planning decisions related to shorter-term operational issues or to longer-term methodological or infrastructural changes. 15

27 3.2 Data 3.2a Precipitation Data Time series of mean monthly precipitation data for all contiguous United States climate divisions (n=344) were downloaded from the National Centers for Environmental Information (NCEI) (ftp://ftp.ncdc.noaa.gov/pub/data/cirs/climdiv/) for the time period from January 1895 through December These data were already quality controlled by NCEI and the records were complete, with no missing data. For this study, climate division data based on the recently developed version two divisional product were used. The new divisional dataset aims to provide more robust divisional averages by accounting for station elevation differences across a division by using interpolation techniques and by using an increased number of stations (Vose et al. 2014). For each climate variable (i.e., temperature, precipitation) the value for each division, year, and month was computed as the area-weighted average of the composite gridpoint values whose centroids fell within the boundaries of that division (Vose et al. 2014). The improved dataset has been quality assured and bias adjustments have been made to ensure that the data are of a high accuracy (Vose et al. 2014). As mentioned previously, the area of study for this research is the contiguous United States, but this large area of study will be examined at a finer resolution using climate regions defined by NCEI (Figure 1). Monthly precipitation values were computed for the nine climate regions by area weighting averages of monthly climate division precipitation data. The area-weighted averages were computed using the Albers equalarea conic projection to determine the area of each climate division. The area of each division was then divided by the total area of the climate region within which that division was located, and the areal weight was then multiplied by the monthly precipitation value for that specific climate division. Monthly time series of area weighted regional precipitation values for each of 16

28 the nine climate regions formed the basis for the precipitation reversal computations and their subsequent analysis within this research. For the remainder of this thesis, the climate regions will be abbreviated as outlined in Table 1. Figure 2: Same as Figure 1 The nine climate regions of the United States as defined by the United States National Centers for Environmental Information (NCEI 2015). The figure is reproduced here in support of the accompanying text and table. Table 1: Abbreviations for the nine climate regions that will be used for the remainder of this thesis. CR1 Northwest CR3 Northern Rockies and Plains CR5 Ohio Valley CR7 Southwest CR9 Northeast CR2 West CR4 Upper Midwest CR6 South CR8 Southeast In order to determine the relative importance of a particular time step for a CR, a monthly precipitation climatology was calculated (Figure 3). The climatology was calculated by determining the average percent of annual precipitation that occurs during each specific month, 17

29 for all CRs. The monthly precipitation climatology reveals that several CRs, including CR1 (Northwest) and CR2 (West), have pronounced dry seasons during the months of June, July, August, and September (Figure 3). CR1 (Northwest) and CR2 (West) also have pronounced wet seasons during the winter and spring months, while spring and summer are the wet periods for CRs 3 (Northern Rockies and Plains) and 4 (Upper Midwest). The annual precipitation is more evenly distributed through the year for the eastern and southern climate regions (Figure 3). 3.2b Niño 3.4 Data A secondary objective of this thesis is to determine if a relationship exists between the El Niño-Southern Oscillation (ENSO) phenomenon and precipitation reversals. The Niño 3.4 index represents an average of sea surface temperatures (SST) across the area of the Pacific Ocean from 5 S to 5 N and from 120 W to 170 W, and it is viewed as a representation of the SST dimension of the ENSO phenomenon. The National Oceanic and Atmospheric Administration (NOAA) defines an El Niño event as when the SST anomalies across the Niño 3.4 region are greater than or equal to 0.5 C for three consecutive months. A La Niña event is defined by three consecutive months of SST anomalies less than or equal to -0.5 C. This study uses the NOAA definitions to define Niño/Niña events, with all else considered to be neither, or neutral. The monthly values of Niño 3.4 were downloaded from January December 2013 ( SOURCES/.Indices/.nino/.EXTENDED/.NINO34/). 18

30 Figure 3: Average percent of precipitation for each month (y-axis), for all nine Climate Regions. The Climate regions are displayed in ascending order (1 to 9) from top to bottom. 19

31 3.3 Methods Extreme Precipitation Reversal (EPR) Identification In order to identify extreme precipitation reversal (EPR) events within the historical records of the nine climate regions, temporal increments were established and PR time series were calculated. According to Goodrich and Ellis (2008), PR values greater than 60 are large enough to likely be of importance to stakeholders. Following this standard, a PR value greater than or equal to 60 will be referred to as an extreme precipitation reversal (EPR), with the magnitude of the reversal listed after the EPR abbreviation (e.g., a reversal of 70 percentiles will be referred as an EPR70 event). For this study, the precipitation data were examined on seasonal and annual timescales. The seasonal analysis was based on four seasons: winter December of the previous calendar year through February; spring March through May; summer June through August; fall September through November. For example, the winter of 1896 included December 1895, January 1896, and February As the record for the winter season was therefore shortened by one year, and a consistent record length was desired, analysis of all seasons was confined to the 119-year period from the winter of 1896 through the fall of For the annual PR events, a year was defined as October 1 through September 30, which is the water year followed by water resource managers, with the benefit being that it does not interrupt the often hydroclimatically critical winter season (as a calendar year does). The annual PR analyses only included 119 years of data for this reason. October 1895 September 1896 was the first year included for analysis in this research, and will be referred to as 1896 throughout this thesis. The years thereafter were also named in a similar manner, with October, November, and December being referred to as part of the following year. At each of the respective time scales, the data were then ranked and a percentile was assigned for each increment of data. The PR value for each increment of data was then calculated 20

32 by subtracting the percentile for the previous record (i.e., season or year) from that for the current record to which the PR value was tagged. This replicated the method used by Goodrich and Ellis (2008). For example, a 2014 year-to-year PR value represents the 2013 annual percentile subtracted from the 2014 annual percentile. Initially, the sign of the reversal (e.g., wetto-dry or dry-to-wet) was not of concern, so the absolute value of the differences was taken. These methods created 118 PR values for the seasonal and annual time scales. These PR lengths were the total lengths of the respective time series data minus one, stemming from the differencing method used. For the EPR frequency results to be statistically valid, it must be revealed that the PR values are random with minimal lag-1 autocorrelation (Goodrich and Ellis 2008). This simply means that every temporal increment should have an equal chance of being wet or dry, independent of the precipitation anomaly for the previous increment or increments. For example, if last year was the wettest on record, this year should still have an equal chance of being dry as it does being wet. A serial correlation analysis was performed for each of the PR time series. For example, the serial correlation analysis for the winter-to-spring comparison examined the winter PR time series and the spring PR time series for correlation. The next step was to determine the expected return frequency and the frequency per century for each classification of EPR events and for each temporal increment. This was done by using a random number generator to create 1000 sets of 119 random numbers, as done by Goodrich and Ellis (2008). The PR methodology was performed on the 1000 sets of data to determine the expected frequency per century and the expected return interval for each EPR class (e.g., EPR90). These values created a baseline to which the actual EPR frequencies for each temporal increment were compared. In order to 21